❯ OpenAI announces three AI research results and invites independent academic scrutiny

Three results enter examinationOn October 26, OpenAI announced model-generated results concerning local smoothing and the Kakeya conjecture, the Exponential Time Hypothesis, and existence questions in four-dimensional quantum field theory. It presents the material as raw research data for academic understanding and examination, with different release approaches for each field.

The complexity claimThe Exponential Time Hypothesis concerns limits on solving certain Boolean satisfiability problems and underpins conditional hardness results elsewhere. OpenAI says its model found a disproof and released a natural-language manuscript and Lean formalization. Lean checks logical steps mechanically. Outside experts, including the hypothesis’s proposers, will organize further scrutiny and explanation. The claim does not directly establish P=NP or an immediately practical fast algorithm.

Release practices are part of the experimentThe mathematics material is public; computer-science material is public with an organized collaboration; physics artifacts remain with an external group for preparation. The physics claim includes Yang–Mills existence. That alone should not be presented as resolving the full Millennium Problem, which also requires a positive mass gap: a strictly positive energy separation between the vacuum and the lowest excitation.

▮ SIGNALAI research is testing both proof generation and the institutions needed to turn machine output into independently understandable, checkable and reusable knowledge.